ExpMaxML

It is an ML algorithm using Expectation Maximization using Gaussian Mixture Models.

Overview

• Given a Dataset, and selected attribute, algorithm clusters and gives the mean for each clusters.
• It uses Gaussian Distribution for finding the probability.
• It uses Expectation Maximization along with log likelihood to find a mean.

Dataset Used

• Indian Liver Test : https://www.kaggle.com/uciml/indian-liver-patient-records
• Wine Quality : https://www.kaggle.com/uciml/red-wine-quality-cortez-et-al-2009

Algorithm

Created using Draw.io

Installation

1. Clone the Repository or Download the Project
2. Navigate to folder ExpMaxML
3. Execute 'python ExpMaxML.py'

Tested Attributes

Liver

Albumin - Very well Tested

Alamine_Aminotransferase

Total_Bilirubin

Wine

fixed Acidity - Very well Tested

Free Sulphur dioxide

Residual Sugar

Custom - Experimental

Sample Execution

1. Select DataSet

Expectation Maximization using Gaussian Mixture Model
1.Liver
2.Wine
3.Custom(File Path Needed)
Select one Dataset from above : 1

2. Select Column

Available Columns
0  Age
1  Total_Bilirubin
2  Direct_Bilirubin
3  Alkaline_Phosphotase
4  Alamine_Aminotransferase
5  Aspartate_Aminotransferase
6  Total_Protiens
7  Albumin
8  Albumin_and_Globulin_Ratio
9  Dataset
Select one Column(Enter number): 7

3. Show Histogram and Find Optimal Cluster

Close the Graph to continue.
Checking feasibility of Cluster:2
Checking feasibility of Cluster:3
Optimal Cluster is 3

4. EM Iteration using GMM

*******  Iteration 0   *********
Printing Mean
2.355348573822168, 3.1193685474989397, 3.9317409062682755
Log Likelyhood :-151537.16537436945
*******  Iteration 1   *********
Printing Mean
2.356583558295428, 3.117896594822594, 3.928358017839534
Log Likelyhood :-151426.94519112387
0.0
*******  Iteration 2   *********
Printing Mean
2.356504406993411, 3.117180941618275, 3.9271702188672144
Log Likelyhood :-151399.42224758756
0.7502912558674304
..
...
.....

5. Final Result and Likelihood Graph

Final Mean Values for each cluster:
Cluster 0  : 2.356364851826904
Cluster 1  : 3.116288502073406
Cluster 2  : 3.925730588575203

Project Struture

ExpMaxML

• ExpMaxML.py - Main Startup File.
• /StatFunctions
  • EM - Methods related to Expectation Maximization
  • PDF - Distribution Implemenataion
  • StatObj - Custom Stat Class

Notebooks

• Liver Analysis - Very Basic Liver Analysis

DataSet

• indian_liver_patient.csv
• winequality-red.csv